Korea Kept Free Allocation at 100%. It Tightened the Benchmark Instead.

The political fight over an emissions trading scheme is almost always about free allocation. How much of it, for whom, and for how long. It is the number trade associations lobby on and the number the press reports.

South Korea just settled that fight in the most generous way available, and then changed something else.

On 11 November 2025 the government approved the National Emission Allowance Allocation Plan for Phase 4 of the Korean Emissions Trading Scheme, covering 2026 to 2030. Sectors judged to be at risk of carbon leakage, including steel, non-ferrous metals, petrochemicals, cement and semiconductors, continue to receive 100% free allocation. On the headline number, industry got what it asked for.

The tightening went into the formula underneath.

What Phase 4 actually is

K-ETS is unusually broad by coverage. It covered 77.75% of Korea’s greenhouse gas emissions in 2023, across 813 entities as of 2025. Korea is now entering its fourth allocation period; Japan’s GX-ETS became mandatory only in April 2026.

Phase 4 sets a total cap of 2,537.3 million tonnes of CO2 equivalent across the five years, including reserves. Around 2.1 billion allowances are distributed free, 260 million are auctioned, and 85 million sit in a quantity-based market stability reserve. Excluding the reserves, roughly 11% of allowances are auctioned.

Eleven percent is not a dramatic number, and read on its own it supports the easy conclusion that Korea blinked. It is the wrong place to look, because the instrument moved somewhere else.

Free allocation is a percentage of something

“100% free allocation” is a statement about a share, not about a quantity. The quantity depends entirely on what the 100% is calculated against, and Korea changed that calculation in three ways at once.

First, the reach of benchmarking. In Phase 3, benchmark-based allocation was to reach 60% of allocation, covering 12 sub-sectors. In Phase 4 the share is to reach 77%, across 15 sub-sectors.

Second, who is newly inside it. Benchmarking is extended to semiconductors, displays and non-ferrous metals, and to fuel use across all industries. For Korea’s single most important export sector, allocation stops being a function of what it emitted and becomes a function of how its intensity compares with its peers.

Third, and most consequentially, the benchmark coefficient tightens from the top 37% of performers to the top 20%.

The third change does the most work. A benchmark allocates against the performance of the best operators in a sector, so moving the definition of “best” from the top 37% to the top 20% leaves every plant below that line with less, without a single percentage point of free allocation being formally withdrawn. The concession and the squeeze sit in different places, and only one of them is in the headline.

Where the auction pressure lands

Auctioning in Phase 4 is split sharply by sector. The power generation sector’s auctioning share rises gradually to 50% by 2030. Other non-power sectors face a 15% auctioning rate, and the carbon-leakage sectors named above sit outside that, at 100% free.

So the paid half of the scheme is aimed at power generation, and the benchmark half is aimed at industry. A steelmaker or a chip fabricator does not buy many allowances at auction. It receives a free allocation computed against a yardstick that now rewards only the top fifth of its sector, and it buys or sells the difference in the secondary market.

This is a more precise instrument than it looks. An auction charges everyone in a sector the same price per tonne regardless of how efficiently they operate. A tightened benchmark charges the laggards and pays the leaders, and it does so without the political cost of visibly taking free allowances away from an export industry. Whether that is a feature or an evasion depends on where the benchmark is set, and the sector-by-sector benchmark values have not been published.

What a covered company has to do about it

The practical burden lands in three places.

The data gets more granular. A grandfathered allocation can be computed from company-level emissions. A benchmark allocation cannot. It needs activity data, output or throughput, at the level the benchmark is defined on, which is usually the process or the product rather than the legal entity. A semiconductor or display manufacturer moving onto benchmarking for the first time is being asked for a different shape of number than the one it has been reporting.

Fuel use is now benchmarked across all industries, which means the question reaches companies that did not consider themselves benchmarked at all. A firm in a sub-sector that remains on grandfathering still faces a benchmark on its fuel.

And the number that matters is now relative. Under grandfathering, a company’s allocation is a function of its own history, and it can forecast it from its own records. Under a benchmark set against the top 20% of a sector, allocation depends on what competitors do. Efficiency improvements by other firms reduce your allocation even if nothing in your own plant changes.

A historical emissions model cannot answer that. A five-year capital plan built on the assumption that last year’s allocation is a reasonable guide to next year’s is working from the wrong variable. The right one is the distance between your intensity and the twentieth percentile of your sector, which is a number most companies do not currently track and cannot derive from their own data alone.

It also changes what an abatement project is worth. Under grandfathering, cutting a tonne saves the cost of a tonne. Under a tightening benchmark, cutting a tonne also moves you up the distribution, and the value of that depends on how crowded the distribution is around you. In a sector where most plants cluster near the same intensity, small improvements move a lot of allocation.

The benchmark values themselves are the live question, and they are set through a process that industry associations participate in. For a company newly inside benchmarking, that process is the only input it can still influence.

The regional picture

Korea and Japan are converging on the same design question from different starting points.

Korea is in its fourth phase and tightening the comparison. Japan’s GX-ETS went mandatory in April 2026, with free allocation under benchmarking in energy-intensive sectors and grandfathering elsewhere, both anchored to a FY2023 to FY2025 baseline that was finished before the obligation began.

Both keep substantial free allocation for trade-exposed industry. Neither can keep it indefinitely, because the EU’s carbon border adjustment mechanism prices the gap between what an exporter paid at home and what an EU producer paid. A tightened benchmark is one way to close that gap on paper without an auction, and it is worth watching whether Korea’s Phase 4 design is read in Brussels as a real carbon cost or as free allocation under another name.

For companies, the sequencing matters more than the politics. The allocation rules of both schemes are now written against activity data at process level, verified, going back several years. A group operating in both jurisdictions is being asked for the same shape of evidence twice, on different formulas, with different baselines.

The number a company can still change is the benchmark. The data behind it is already history.

Tech for Impact Summit 2027 runs on 18 and 19 May 2027 at the United Nations University in Tokyo, co-hosted by UNU. Senior leaders from business, policy and capital spend two days on questions of this kind. Participation is by invitation. If this is the work you do, ask us for one at tech4impactsummit.com/apply.